论文部分内容阅读
短期地震概率(STEP)预测模型将大森—宇津余震衰减关系和古登堡—里克特频度—震级关系应用到地震丛集。这个模型主要用来预测余震活动,而依赖于时间不变的背景模型预测大多数主震。另一方面,长期地震预测模型EEPAS(根据尺度每个地震都有前兆)利用前兆尺度增加现象及相关的预测标度关系,取决于震级的大小可提前数月、数年或数十年预测主震。这两种模型均显现出比地震活动性不随时间变化的模型可提供更多的信息。通过将这两种模型混合在一起,我们期望形成包含更多信息的短期预测模型。本优化混合模型利用了加利福尼亚州1984～2004年间改进的国家地震系统目录,是一种预测M≥5.0级地震的凸线性组合,其中EEPAS预测量为0.42,STEP预测量为0.58。这种混合模型与每一种单独模型相比,平均概率增益大于2。数种不同的混合模型将提交到南加州地震中心的地震预测能力研究合作实验室(CSEP)检测中心,用以确定该结果是否由这些模型对未来地震的实时检测产生出。
Short-Term Earthquake Probability (STEP) Prediction Model The relationship between the attenuation of the Omori-Ujin aftershock and the Gutenberg-Richet frequency-magnitude relationship is applied to the earthquake cluster. This model is mainly used to predict aftershock activity, but relies on a time-invariant background model to predict most of the main shocks. On the other hand, the long-term earthquake prediction model EEPAS (with precursors for each earthquake on a scale basis) exploits the increase of precursory scales and the associated predictive scaling relationships depending on the size of the magnitude that can be predicted many months, years or decades ahead of schedule shock. Both models show more information than models whose seismicity does not change with time. By mixing these two models together, we expect to form a short-term forecasting model that contains more information. This optimized hybrid model uses a catalog of improved national seismological systems from 1984 to 2004 in California as a convex linear combination predicting M ≥ 5.0 earthquakes with an EEPAS forecast of 0.42 and a STEP forecast of 0.58. The average probability gain for this hybrid model is greater than two for each individual model. Several different hybrid models will be submitted to the CSEP Center for Earthquake Prediction Capability Testing at the Southern California Earthquake Center to determine if the results were generated by these models for real-time detection of future earthquakes.